Myristoylated alanine-rich C kinase substrate (MARCKS) sequesters spin-labeled phosphatidylinositol 4,5-bisphosphate in lipid bilayers.

The myristoylated alanine-rich protein kinase C substrate (MARCKS) may function to sequester phosphoinositides within the plane of the bilayer. To characterize this interaction with phosphatidylinositol 4,5-bisphosphate (PI(4,5)P(2)), a novel spin-labeled derivative, proxyl-PIP(2), was synthesized and characterized. In the presence of molecules known to bind PI(4,5)P(2) the EPR spectrum of this label exhibits an increase in line width because of a decrease in label dynamics, and titration of this probe with neomycin yields the expected 1:1 stoichiometry. Thus, this probe can be used to quantitate the interactions made by the PI(4,5)P(2) head group within the bilayer. In the presence of a peptide comprising the effector domain of MARCKS the EPR spectrum broadens, but the changes in line shape are modulated by both changes in label correlation time and spin-spin interactions. This result indicates that at least some proxyl-PIP(2) are in close proximity when bound to MARCKS and that MARCKS associates with multiple PI(4,5)P(2) molecules. Titration of the proxyl-PIP(2) EPR signal by the MARCKS-derived peptide also suggests that multiple PI(4,5)P(2) molecules interact with MARCKS. Site-directed spin labeling of this peptide shows that the position and conformation of this protein segment at the membrane interface are not altered significantly by binding to PI(4,5)P(2). These data are consistent with the hypothesis that MARCKS functions to sequester multiple PI(4,5)P(2) molecules within the plane of the membrane as a result of interactions that are driven by electrostatic forces.

brane, and it plays a key role in cell signaling. PI(4,5)P 2 a substrate for phospholipase Cs and the precursor for the second messengers inositol trisphosphate and diacylglycerol (1). In addition, PI(4,5)P 2 is itself an important signaling molecule (2). PI(4,5)P 2 regulates cell motility and morphogenesis thorough its interaction with PH domains and actin-binding proteins (3)(4)(5) and plays a critical role in membrane trafficking (6,7), the regulation of ion channel activity (8), and the activation of guanine nucleotide exchange (9). Although not as abundant as the major phospholipids phosphatidylcholine or phosphatidylserine, PI(4,5)P 2 appears to be present in the plasma membrane at roughly constant levels on the order of 1 mol% (3). At these levels of PI(4,5)P 2 , there would be large numbers of these lipids available to interact with proteins, and there is considerable interest in understanding how these interactions are regulated within the bilayer.
One concept that has emerged to explain the specific action of PI(4,5)P 2 is that local free levels of PI(4,5)P 2 within the bilayer are controlled by modulating its lateral heterogeneity. There are several ways in which the distribution of PI(4,5)P 2 within the cell might be regulated. Clearly, the enzymes that make PI(4,5)P 2 may be sequestered in certain regions of the bilayer, or PI(4,5)P 2 may be specifically accumulated into specialized lipid domains such as cholesterol-rich lipid rafts (10). Proteins might also sequester or mask the presence of free PI(4,5)P 2 , and there is evidence that the myristoylated alaninerich protein kinase C substrate (MARCKS) specifically binds and regulates the levels of free PI(4,5)P 2 within the bilayer (11)(12)(13)(14). MARCKS is a Ca 2ϩ -dependent protein kinase C substrate that is present in high concentrations in many cell types (15), it associates with the membrane in-terface though its N-terminal myristoylation as well as an electrostatic interaction of its highly basic effector domain with the membrane interface (16). The MARCKS effector domain (resi-dues 151-175) is highly basic having the sequence KKKKKRFSFKKS-FKLSGFSFKKNKK. MARCKS binds strongly to Ca 2ϩ -calmodulin through this domain (17,18), and this domain interacts with actin (19). Membrane-bound MARCKS may be dissociated from the membrane interface by calmodulin and by protein kinase C, which phosphorylates MARCKS within its effector domain (20).
Although it has been implicated in a number of cellular processes, the exact role of MARCKS has been unclear. Because phosphorylation prevents MARCKS from binding to calmodulin, it was suggested that MARCKS might allow for an interaction between protein kinase C and calmodulin-dependent signaling pathways (17); however, recent work suggests that MARCKS actually functions to sequester PI(4,5)P 2 within the plane of the bilayer. The binding of MARCKS (151-175) to the membrane interface has been shown to inhibit the hydrol-ysis of PI(4,5)P 2 by either phospholipase C (PLC)-␦ or PLC-␤ (11,14), presumably because this peptide competes successfully with the active site of PLC for PI(4,5)P 2 . MARCKS has been shown to accumulate at lipid rafts and to codistribute with PI(4,5)P 2 (13), and a peptide derived from the effector domain of MARCKS (MARCKS (151-175)) has been shown to bind strongly to membranes containing PI(4,5)P 2 and PI(3,4)P 2 (12,14). As a result of this interaction, MARCKS may bind a significant fraction of PI(4,5)P 2 within the plasma membrane. Free levels of PI(4,5)P 2 in the membrane could then be controlled either through protein kinase C, by controlling the phosphorylation state of the MARCKS effector domain, or through the cytoplasmic levels of Ca 2ϩ -calmodulin.
In the present study, we examined the interaction between MARCKS (151-175) and PI(4,5)P 2 using two approaches. First, we synthesized a novel spin-labeled derivative of PI(4,5)P 2 in which a proxyl nitroxide spin label was incorporated into the sn-1 acyl chain of the phospholipid (21). EPR spectroscopy then was used to examine the interactions between this spin label and known PI(4,5)P 2 -binding macromolecules and the interactions between this label and MARCKS (151-175). Second, we derivatized a series of cysteine-substituted peptides based on MARCKS (151-175) with a sulfhydryl reactive spin label ( Fig.  1) and used EPR spectroscopy to examine the conformation and structure of MARCKS (151-175) in the presence and absence of PI(4,5)P 2 . The data demonstrate that the proxyl-PIP 2 is a useful probe for the interactions between PI(4,5)P 2 and molecules that bind the PI(4,5)P 2 head group within the plane of the bilayer. The data also indicate that MARCKS (151-175) associates with multiple PI(4,5)P 2 molecules within the plane of the membrane by a process that is driven largely by electrostatic interactions.

Materials
Palmitoyloleoylphosphatidylserine (PS), palmitoyloleoylphosphatidylcholine (PC), 5-doxyl phosphatidylcholine (5-doxyl PC), and the ammonium salt of phosphatidylinositol 4,5-bisphosphate (PIP 2 ) were obtained from Avanti Polar Lipids (Alabaster, AL). Neomycin was obtained from Calbiochem. Spin-labeled derivatives of MARCKS (151-175) were synthesized and purified by HPLC as described previously (22) or obtained from the Biomolecular research facility at the University of Virginia. The identity of the labeled peptides was confirmed by mass spectrometry, and they had a purity in excess of 97% as determined by HPLC and in-line detection by UV spectroscopy of the peptide backbone at 210 nm.

Methods
Synthesis of a Spin-labeled Derivative of PI(4,5)P 2 (Fig. 2) N-Hydroxysuccinimydyl-3-carboxylate Proxyl, Free Radical (1)-A suspension of 3-carboxy proxyl, free radical (99.9 mg, 0.54 mmol) and N-hydroxysuccinimide (67.6 mg, 0.59 mmol) was prepared in 10 ml of CH 2 Cl 2 (distilled from CaH 2 ) and stirred for 20 min. Dicyclohexylcarbodiimide (116.1 mg, 0.56 mmol) and dimethylaminopyridine (24.0 mg, 0.20 mmol) were added, and the reaction mixture was stirred overnight at room temperature. The precipitate was filtered off, and the filtrate was evaporated to dryness. The residue was suspended in EtOAc, filtered, and the filtrate evaporated to dryness. The product was purified through a plug of silica, eluted with 3:2 hexanes:EtOAc. The yield was 91.7 mg (60%) as an orange oil. This product yielded a single spot by TLC,  7.8) and stirred overnight at room temperature. The reaction mixture was concentrated to dryness, and the residue was washed with acetone (5 ϫ 1.5 ml). The crude product was dissolved in 2 ml of water and applied to a small column (12 ϫ 15 mm) of DEAE-cellulose. The column was eluted with a step gradient (0.2 M steps) of 0.2-2.0 M TEAB (2-ml portions) and finally with 3:7 MeOH:TEAB (2 M). The product started to elute with 1.6 M TEAB and finished with the MeOH: TEAB mixture. The desired fractions, detected by phosphate assay, were pooled and lyophilized yielding the product as the triethylammonium salt. The yield was 2.7 mg. ES-MS: m/z ϭ 1015. 3 [MϩH] ϩ . The product ion was the only major peak in the mass spectrum, and no ion corresponding to the mass of the starting material, 2, was detected. Furthermore, the product was tested with a ninhydrin assay, and no amine was found to be present, indicating the absence of the starting material, 2. EPR spectroscopy indicated that there was no detectable unreacted proxyl-nitroxide present in the sample, and the EPR spectra of this product in CHCl 3 and bound to membranes was consistent with that expected for a spin-labeled amphiphile. The presence of a significant non-spin-labeled species in the purified material was ruled out by comparing the concentration of 3 determined gravimetrically with that obtained by double integration of the EPR signal of 3. These concentrations were in agreement, within experimental error, and were also in agreement with the concentration determined from total phosphate (24). Furthermore, titration of the product with neomycin (see "Results") indicated a 1:1 binding with the appropriate affinity. These assays indicated that the purity of 3 was a minimum of 90%.

Expression and Purification of PLC-␦1 PH Domain
Recombinant PH domain from human PLC-␦1 was produced from an Escherichia coli strain that was generously provided by Mario Rebecchi and purified following a procedure described previously (25). The identity of the PH domain was confirmed by gel electrophoresis and mass spectrometry, and the purity as judged by electrophoresis was greater than 95%.

Lipid Vesicle Preparation
Lipid bilayers having the desired lipid composition were produced by mixing the appropriate lipids from stock solutions in chloroform, removing the chloroform by vacuum desiccation, and hydrating the resulting lipid film in a buffer containing 100 mM KCl, 10 mM MOPS, pH 7.0. The lipid mixture was freeze-thawed five times, and unilamellar vesicles were produced by extrusion of the mixture through 1,000 Å polycarbonate filters (Poretics, Livermore, CA) using a LiposoFast extruder (Avestine, Ottawa, Canada). Proxyl-PIP 2 could be incorporated into both leaflets of the membrane by dissolving the labeled lipid into the lipid chloroform solution; however, in all of the data shown here, the spin label was incorporated into the outer membrane leaflet by adding the vesicle solution to a dried film of proxyl-PIP 2 . The addition of unlabeled or spin-labeled MARCKS (151-175) to preformed lipid vesi-

EPR Spectroscopy
EPR spectra for either proxyl-PIP 2 or spin-labeled MARCKS (151-175) were obtained at X-band from ϳ5 l of sample using a Varian E-line century series spectrometer fitted with a MITEQ microwave amplifier (Hauppauge, NY) and a two-loop one-gap resonator (Medical Advances, Milwaukee, WI). Nonsaturated EPR spectra were obtained using a microwave power of ϳ2 mW or less and a modulation amplitude of 1 gauss peak to peak.
EPR spectroscopy was used to titrate the interaction between proxyl-PIP 2 and MARCKS (151-175) or neomycin using an approach similar to that described previously (26). In this case, 50 -100 l of a lipid vesicle suspension at a lipid concentration of 20 mM having 0.25-0.5% proxyl-PIP 2 was titrated with neomycin or MARCKS (151-175) by measuring the first derivative peak-to-peak (or peak-to-trough) amplitude of the central proxyl nitroxide resonance, A(0). Using a stainless steel plunger, ϳ10 l of sample was drawn into or removed from a (0.5-mm inner diameter ϫ 0.7-mm outer diameter) quartz capillary (VitroCom, Mt. Lakes, NJ) that was fitted into the loop-gap resonator. The amplitude of the central nitroxide resonance, A(0), was then recorded as a function of the concentration of neomycin or MARCKS (151-175) added to the lipid sample. The EPR spectrum of proxyl-PIP 2 in the presence of neomycin or MARCKS (151-175) is a linear combination of EPR signals from free lipid and lipid bound to the macromolecule. If the line shapes for these species is known, the fraction of bound proxyl-PIP 2 may then be determined from the value of A(0) (see Equation 6).

EPR Power Saturation Measurements
Power saturation measurements were made on proxyl-PIP 2 and on spin-labeled MARCKS (151-175) to determine the depth of the nitroxide label from the level of the lipid phosphate. These measurements were carried out in a manner similar to that described previously (27) using gas-permeable TPX capillary tubes (Medical Advances, Milwaukee WI). In each case, the peak-to-peak (or peak-to-trough) amplitude of the central (m I ϭ 0) first derivative EPR resonance A(0) on the incident microwave power, P, was measured and fit to the expression where I is a scaling factor, P 1/2 is the microwave power required to reduce the resonance amplitude to half its unsaturated value, and ⑀ is a measure of the homogeneity of the saturation of the resonance (28). P 1/2 was determined with I, ⑀, and P 1/2 as adjustable parameters for each label under three conditions: when equilibrated with a N 2 , equilibrated with air (20% O 2 ), or equilibrated with N 2 in the presence of 20 mM NiEDDA (in PC membranes, NiAA was used as the paramagnetic metal). The value of ⌬P 1/2 oxy or ⌬P 1/2 NiEDDA was determined from the difference in P 1/2 in the presence and absence of either NiEDDA or O 2 , respectively. For each sample a depth parameter, ⌽, was determined from the values of ⌬P 1/2 according to Equation 2.
The parameter ⌽ is directly related to the difference in the standard state chemical potentials of O 2 and NiEDDA, which vary as a function of depth in the lipid bilayer. As a result, ⌽ provides an estimate of the nitroxide depth in the lipid bilayer (28).

Analysis of Binding Data
The 1:1 binding of neomycin or other macromolecule (M) to proxyl-PIP 2 was analyzed in an manner similar to that described previously (14) for the equilibrium.
The apparent association constant, K a , for 1:1 binding is given by Equation 3, The solution of Equations 3-5 yields a quadratic that can be used to predict the 1:1 binding as a function of the concentration of neomycin or other PIP 2 -binding species. This expression can be used to predict the amplitude of the central EPR resonance amplitude, A(0), as a function of added macromolecule. As indicated above, the EPR spectrum is a simple sum of EPR spectra from the free and bound proxyl-PIP 2 , as a result the magnitude of A(0) may be written as where A f and A b represent the intrinsic amplitudes of free and macromolecule-associated proxyl-PIP 2 . Equation 6 was used in combination with Equations 3-5 to determine the 1:1 binding behavior of proxyl-PIP 2 .
Using a similar approach, we also analyzed the titration of proxyl-PIP 2 with MARCKS (151-175) for the case where multiple PIP 2 bind to the effector domain of MARCKS. In this case we assume the equilibrium in Reaction 2.
In this equilibrium it is assumed that MARCKS (151-175) binds in a single step to n proxyl-PIP 2 , and a quadratic equation that describes this binding may be derived using equations analogous to 3, 4, and 5 shown above. It should be noted that this binding assumes that n proxyl-PIP 2 are in a preformed complex before binding to MARCKS (151-175). This is almost certainly not the case, and Reaction 2 should be taken only as an approximation of the actual events that take place during peptide binding. Fig. 3A is an EPR spectrum of the triethylammonium salt of proxyl-PIP 2 dissolved into chloroform at a concentration of ϳ200 M. The EPR spectrum exhibits clear evidence for spin exchange, which likely results from the formation of inverted micelles in this nonpolar solvent. In the aqueous phase, PI(4,5)P 2 is known to form micelles (29), and this also appears to be the case for proxyl-PIP 2 . Although proxyl-PIP 2 is freely soluble in aqueous solution, no EPR spectrum can be observed at room temperature. This is consistent with the formation of micelles, which would promote strong dipolar interactions and/or spin exchange between labeled nitroxides and result in high relaxation rates. The lack of an observable EPR spectrum also indicates that the critical micelle concentration for proxyl-PIP 2 must be lower than 5 M because an aqueous concentration of monomeric proxyl-PIP 2 at or above this concentration would have yielded a well resolved signal. When lipid vesicles formed from PC are added to solid proxyl-PIP 2 as described above (see "Methods"), the EPR spectrum shown in Fig. 3B is observed. This spectrum is characteristic of a label that is monomeric in the membrane and undergoing relatively rapid motion, and it is reasonably well simulated assuming an isotropic rotational model where the nitroxide has a correlation time of about 6 ns.

Proxyl-PIP 2 Incorporates Spontaneously into Lipid Vesicles-Shown in
To ensure that the spectrum in Fig. 3B arises from a nitroxide with a membrane location, the EPR spectrum was power saturated (see "Methods") in the presence of O 2 , NiAA, or NiEDDA. The values of ⌬P 1/2 for O 2 and the metal complexes are consistent with a membrane location for the label. For proxyl-PIP 2 in the presence of PC, we obtain a depth parameter, ⌽, of Ϫ0.77 using 20 mM NiAA. Using a calibration curve determined previously for PC bilayers (28), a location of 3 Ϯ 2 Å below the level of the lipid phosphate is obtained. For proxyl-PIP 2 incorporated into PC:PS membranes, a value of ⌽ ϭ 0.46 is obtained using 20 mM NiEDDA. This value yields a position of 6 Ϯ 2 Å below the level of the lipid phosphate based on a calibration determined recently for PC:PS (30). Thus, the power saturation data are consistent with a membrane location for the proxyl-PIP 2 , where the label takes up a position a few Å within the bilayer below the level of the head group phosphate. This position is ϳ10 Å shallower than that expected if the glycerol backbone of the proxyl-PIP 2 were located at the same position as the membrane lipid, and the acyl chain attached to the proxyl spin label were in a fully extended conformation.
Double integration of the spectrum in Fig. 3B yielded a spin concentration of ϳ90 M, which is close to the concentration of nitroxide label added to the vesicle suspension. Thus, this label fully incorporates into the lipid bilayer when absorbed to vesicles from the external aqueous solution. Because PI(4,5)P 2 has ϳ3 negative charges at neutral pH it is not expected to undergo transmembrane migration; as a result, proxyl-PIP 2 that is incorporated in this manner should reside on the external surface of these preformed lipid vesicles. In subsequent experiments, detailed below, proxyl-PIP 2 has been incorporated in this manner into the external vesicle surface.
Proxyl-PIP 2 Is Sensitive to the Interaction with Neomycin and the PH Domain from PLC-␦1-To determine whether the EPR spectrum of proxyl-PIP 2 is sensitive to interactions of its head group, we compared the EPR spectra of proxyl-PIP 2 in the presence and absence of neomycin and the PH domain from PLC-␦1, two well characterized PI(4,5)P 2 -binding molecules. Shown in Fig. 4, A and B, are EPR spectra for proxyl-PIP 2 in the presence and absence of neomycin and the PLC-␦1 PH domain. Both of these molecules are known to exhibit strong 1:1 binding to PIP 2 , and in both cases, the EPR spectra exhibit a similar broadening in the presence of these reagents. This In these samples, proxyl-PIP 2 was at a concentration of 0.5 mol%, and the total PC concentration was ϳ20 mM. Each pair of spectra was recorded under identical conditions and with identical spin concentrations so that their amplitudes would be normalized. level of broadening corresponds to an increase in the rotational correlation time for the proxyl label of about 25%.
There are several mechanisms that might give rise to this change in rotational correlation time. The proxyl spin label is quite mobile when attached to PI(4,5)P 2 , which is not unreasonable given the alkyl chain and rotatable bonds that link the label to the glycerol backbone. However, bond rotations and librations of this chain are expected to be of limited amplitude, and some averaging of the nitroxide magnetic interactions should result from lipid rotational motion that takes place on the ns time scale (31). As a result, any interaction that slows the rotational rate of the labeled lipid, such as attachment to a large macromolecule, should broaden its EPR spectrum. Conceivably, interactions with the PI(4,5)P 2 head group might sterically interfere with the spin label and reduce the amplitude of motion of the proxyl label attached to the sn-1 chain. Finally, the if the interaction with proxyl-PIP 2 altered the membrane position of the label, changes in label motion might result because of the altered environment around the label.
To determine whether more quantitative information can be extracted from the interaction of proxyl-PIP 2 with a PI(4,5)P 2binding species, we titrated the first derivative EPR line amplitude (see "Methods") as a function of the neomycin concentration. Shown in Fig. 5 are the first derivative EPR amplitudes obtained from this titration. Addition of neomycin results in the formation of a PI(4,5)P 2 ⅐neomycin complex and a change in label dynamics. These data points were then fit using Equations 3-6 assuming a 1:1 binding of neomycin to PIP 2 . In this fit, both K a and A b were allowed to be adjustable parameters. The agreement between these data and this fit is quite reasonable, and the accuracy of this fit is very sensitive to the stoichiometry of the interaction between PIP 2 and neomycin. The value of the molar partition coefficient, K a , which is obtained in this fit is ϳ3 ϫ 10 5 M Ϫ1 , which is close to that obtained previously (32).
It should be noted that neomycin has been reported to promote the transport of phosphoinositides across cell membranes (33). The mechanism leading to this transport is not understood, and experiments to determine whether neomycin can facilitate the transport of proxyl-PIP 2 in the model membrane systems used here are currently in progress.
Multiple PI(4,5)P 2 Molecules Interact with MARCKS-Shown in Fig. 6A are EPR spectra of proxyl-PIP 2 in the presence and absence of MARCKS (151-175) in PC vesicles. MARCKS (151-175) has a high affinity for membranes containing PI(4,5)P 2 (14), and in Fig. 6A sufficient peptide has been added to bind all of the available proxyl-PIP 2 completely. As seen previously for neomycin and the PH domain, MARCKS (151-175) also produces a broadening of the proxyl-PIP 2 line shape. However, the broadening that is seen in the presence of MARCKS (151-175) is greater than that seen in Fig. 4 for neomycin or the PH domain.
At least two mechanisms may be acting to produce the line width changes seen in Fig. 6A. First, MARCKS (151-175) may diminish the motional averaging of the proxyl-labeled lipid in a manner similar to that seen for the PH domain or neomycin. Second, if MARCKS (151-175) binds multiple PI(4,5)P 2 molecules, an additional line width broadening might result from the proximity between spin labels. To determine whether some of the broadening is caused by the proximity between spin labels, the spectra shown in Fig. 6B were obtained under conditions where proxyl-PIP 2 was diluted with unlabeled PI(4,5)P 2 . When the spin-labeled lipid is diluted by a factor of 1:3, the line shapes are still broadened upon the addition of MARCKS (151-175), but there is significantly less broadening than was seen in Fig. 6A. The effect of diluting the proxyl-PIP 2 suggests that multiple PI(4,5)P 2 species interact with MARCKS (151-175). If two or more proxyl-PIP 2 are brought sufficiently close to each other so that dipole-dipole or collisional exchange mechanisms take place, an additional line width broadening will result. When these samples are rapidly frozen in LN 2 (data not shown), a dipolar broadened spectrum can be observed. Low temperature eliminates the effects of motion on the nitroxide EPR spectrum and allows the strength of the dipolar interaction to be estimated (34). Assuming a pairwise interaction, the EPR spectra indicate that proxyl-PIP 2 s are separated by distances on the order of 18 -22 Å when bound to MARCKS (151-175) (35).
The EPR spectra shown in Fig. 6A were titrated as a function of the concentration of MARCKS (151-175), and Equation 6 was used to estimate the fraction of bound proxyl-PIP 2 . The result is shown in Fig. 7 along with a curve generated using Equations 3-5 corresponding to a 1:1 binding of proxyl-PIP 2 to MARCKS (151-175). These data cannot be fit to a 1:1 binding, regardless of the choice of affinity constants. Also shown in Fig.  7 is a curve representing the best fit using the equilibrium given by Reaction 2 where both n and K a are allowed to be adjustable parameters. A value of n ϭ 3 produces the best fit to the data, but any stoichiometry (PI(4,5)P 2 :MARCKS) in the range of 2.5-3.5 gives an acceptable fit to these data. The apparent molar partitioning obtained in this fit is 4 ϫ 10 5 M Ϫ1 , somewhat less than that expected based upon the binding measured under dilute conditions (14). The binding of MARCKS (151-175) is known to depend strongly on the surface change density, and this slightly lower apparent affinity is not surprising given that the titration covers a range that saturates the free PIP 2 (and thus the negative surface charge density) in the bilayer. The equilibrium given in Reaction 2 is also highly simplified and cannot represent the actual molecular steps taking place. As a result the fit (solid line) shown in Fig.  7 must be viewed with some caution. Nonetheless, taken together with the data in Fig. 6, this titration provides strong evidence that multiple PI(4,5)P 2 molecules interact with the effector domain of MARCKS.
MARCKS Does Not Change Position or Structure When Bound to PI(4,5)P 2 -containing Membranes-Previous work on MARCKS (151-175) demonstrated that this peptide assumed an extended structure when bound to PC:PS with its five phenylalanine residues buried within the interface (22). We find no evidence for significant structural changes in MARCKS (151-175) when complexed to PI(4,5)P 2 . Shown in Table I are the central line widths for the EPR spectra of five single spinlabeled and five double spin-labeled MARCKS (151-175) in PC:PS (3:1) or PC:PI(4,5)P 2 . In all cases, except one, the EPR line shapes are identical (within experimental error) when bound to PC:PS or PC:PI(4,5)P 2 . There is a slight decrease in line width at a position closest to the N-terminal end of the peptide. This highly charged end of the peptide is positioned off the membrane interface, and the change in line shape may represent a slight shift in its position when these two lipid mixtures are compared. For the double labeled peptides, the increased line widths for i, iϩ3 and i, iϩ7 compared with the single labeled species are caused by interactions between the nitroxides (either dipole-dipole or spin exchange). These line widths should be strongly dependent upon the average separation between nitroxides and hence any change in average shape or secondary structure. These line widths are identical in the presence and absence of PI(4,5)P 2 .
Shown in Fig. 8 are the EPR spectra for spin-labeled MARCKS (151-175) bound either to PC:PS-or PC:PI(4,5)P 2containing membranes. Shown in Fig. 8A are two single labeled spectra for MARCKS (151-175), whereas Fig. 8B shows EPR spectra for two peptides with nitroxide pairs separated by i, iϩ7, and i, iϩ11, respectively. The spectra bound to PC:PS are unchanged compared with the case where the peptide is bound to PC:PI(4,5)P 2 membranes. The single labeled spectra arise from nitroxides that have a correlation time of about 3-4 ns, consistent with their attachment to an extended flexible peptide. In PC:PI(4,5)P 2 , depth measurements were made for several singly spin-labeled MARCKS peptides and compared with depths obtained previously in PC:PS (3:1). The data are shown in Table II. In the central and C-terminal region of the peptide the label is at a depth of ϳ7 Å below the level of the lipid phosphate in PC:PI(4,5)P 2 , and at the N terminus, the label is in the aqueous phase several Å from the lipid phosphate in this lipid mixture. These depths are identical, within experimental error, to the depths obtained previously in PC:PS (22).
Taken together, these data indicate that binding by PI(4,5)P 2 produces no significant conformational change in the membrane-bound structure of the effector domain of MARCKS, consistent with the less direct findings of CD measurements (14); in addition, PI(4,5)P 2 binding does not change the position of the MARCKS domain at the membrane interface. For all of the labeled residues examined, PI(4,5)P 2 binding does not significantly alter the dynamics of the spin-labeled side chain. DISCUSSION The work that is described here was carried out with several objectives in mind. First we wanted to investigate a spinlabeled derivative of PI(4,5)P 2 to determine whether it would provide a probe for protein-polyphosphoinositide head group interactions within the plane of the bilayer. Second, we wanted to characterize the interactions between the effector domain of MARCKS and PI(4,5)P 2 . The proxyl-PIP 2 spin label synthesized here readily dissolves in chloroform or aqueous solution, presumably as inverted or normal micelles as discussed above. As a result, it was possible to incorporate reproducibly the probe into one or both leaflets of a lipid vesicle.
The EPR spectrum of the proxyl-PIP 2 broadens when it binds to either neomycin or the PH domain from PLC-␦1, and the change in line shape appears to be the result of a change in the dynamics of the proxyl label, so that the extent of motion averaging of the magnetic interactions of the nitroxide is reduced. When the EPR signal of proxyl-PIP 2 is titrated with In these samples, phosphatidycholine is present at a concentration of 20 mM, and proxyl-PIP 2 is incorporated into the PC bilayer at a concentration of 0.5 mol%. For each pair of spectra, the amplitudes with and without peptide have been normalized by recording the spectra under identical conditions. neomycin, the expected 1:1 stoichiometry is revealed. Thus, this probe is sensitive to interactions with macromolecules that are known to bind the PI(4,5)P 2 head group. In addition, the probe is sensitive to the local clustering of PI(4,5)P 2 , which gives rise to spin-spin interactions between nitroxide labels. This sensitivity is potentially extremely useful for the examining the lateral heterogeneity in PI(4,5)P 2 , and it is a feature that should prove useful in determining whether PI(4,5)P 2 is enriched in certain types of lipid phase separations or in certain protein-induced lipid domains. Although proxyl-PIP 2 appears to be a good probe for interactions made by the head group of PI(4,5)P 2 , it is not expected to be a good probe for interactions made by the acyl chain moiety of PI(4,5)P 2 . The acyl chain region of proxyl-PIP 2 differs significantly from that of naturally occurring PI(4,5)P 2 , and it does not resemble the diacylglycerol moiety of this lipid.
Power saturation of the EPR spectrum of proxyl-PIP 2 indicates that that the label on this lipid lies near the membrane interface. If the short alkyl chain that links the lipid backbone to the proxyl spin label were in a fully extended form, and the PI(4,5)P 2 glycerol backbone were placed in a position similar to that for PC, the label would be expected to lie at a depth of ϳ15 Å. The position of the label at the interface may be a result of one or a combination of two effects. First, the sn-1 chain might assume a highly bent average configuration that places the proxyl spin label at the membrane interface. Indeed, similar effects have been observed for certain fluorescent labeled lipids that have significant hydrophilic character, e.g. 7-nitrobenz-2oxa-1,3-diazole-labeled lipids (36). Second, the highly charged PI(4,5)P 2 head group might be placed further on the aqueous side of the membrane interface than other membrane lipids. This could arise because of a greater Born repulsion resulting from the three negative charges on the PI(4,5)P 2 head group. Depth measurements on proxyl-PIP 2 show that the spin label is positioned 5 Å deeper into the bilayer in the presence of MARCKS (151-175) (data not shown), suggesting that the position of this labeled lipid along the bilayer normal is variable and sensitive to electrostatic interactions in the interface.
A peptide derived from the effector domain of MARCKS was recently shown to bind strongly to membranes containing PI(4,5)P 2 (12,14). This finding supports the idea that MARCKS functions to sequester polyphosphoinosites within the plane of the membrane. As depicted in Fig. 9, MARCKS interacts with the plasma membrane interface and binds to PI(4,5)P 2 , preventing PI(4,5)P 2 from freely diffusing within the plane of the bilayer. This interaction can be removed and PI(4,5)P 2 released when MARCKS is phosphorylated by protein kinase C or when the concentration of Ca 2ϩ -calmodulin increases. The interaction between PI(4,5)P 2 and the effector domain of MARCKS is revealed clearly in the EPR spectrum of proxyl-PIP 2 . As with neomycin and the PH domain, the interaction appears to slow the rotational rate of the spin label on PI(4,5)P 2 . In addition, the interaction between proxyl-PIP 2 (151-175). The total proxyl-PIP 2 concentration is 50 M in PC vesicles at a lipid concentration of ϳ20 mM. The data shown (q) were obtained from two independent titration experiments. The solid line represents a nonlinear least squares fit through the data according to Reaction 2 and yields a binding stoichiometry of 3:1 (PI(4,5)P 2 : MARCKS). The dashed line represents the predicted binding assuming a 1:1 stoichiometry between PI(4,5)P 2 and MARCKS (151-175) through the data using Equations 3-6. In each case the curves correspond to a value for K a of 4 ϫ 10 5 M Ϫ1 .

TABLE I Central EPR line width (⌬H) for MARCKS-derived peptides (gauss)
The central EPR line width is the peak-to-peak splitting of the first derivative of the m I ϭ 0 resonance. Errors in the line width are approximately Ϯ 0.1 gauss unless otherwise indicated. The peptides indicated were single or double cysteine substitutions of a peptide derived from the MARCKS effector domain having the sequence Ace-KKKKKRF-SFKKSFKLSGFSFKKNKK-NH 2 . These peptides were derivatized with the methanethiosulfonate spin label (Fig. 1) to produce peptides with single or double spin label side chain(s) (R1) at the indicated position(s). gives rise to spin-spin exchange or dipolar broadening, indicating that more than one proxyl-PIP 2 is bound to the effector domain of MARCKS. The data obtained here are consistent with a stoichiometry of about 2.5-3.5 proxyl-PIP 2 /MARCKS, a number that is in approximate agreement with that found based on competition and electrophoretic mobility measurements (14). The association between MARCKS (151-175) and proxyl-PIP 2 is strong enough to slow the rotational rate of the labeled lipid and/or alter the dynamics of the proxyl moiety, but this binding does not appear to involve specific molecular contacts between the lipid and this peptide. This peptide is in a flexible, extended structure when associated with membrane surfaces containing acid lipids such as PS (22), and the spin-labeled R1 side chain is highly sensitive to tertiary contact and to changes in dynamics of the peptide backbone (37); as a result, if specific van der Waals contacts or hydrogen bonds were required for the PI(4,5)P 2 interaction, changes in the EPR line shapes of these spin-labeled derivatives of MARCKS (151-175) would have been seen. The spectra of the double labeled peptides also show no change when binding to PS versus PI(4,5)P 2 membranes. These spectra will also be highly sensitive to the average distances between labels and thus to the average configuration of the backbone.
A likely explanation for the lack of any structural change when MARCKS associates with PI(4,5)P 2 is that this interaction is driven largely by electrostatic interactions. Electrostatics and a free energy contribution for burying phenylalanine side chains within the interface are responsible for the association of MARCKS (151-175) with membranes containing PS (38). The data obtained here indicate that MARCKS (151-175) interacts identically with the membrane interface in the presence of either PS or PI(4,5)P 2 ; thus, the same electrostatic interactions must be important for the association of this peptide to PI(4,5)P 2 . A number of other observations support the conclusion that electrostatics is important. For example, the interaction between MARCKS and PI(4,5)P 2 does not discriminate between PI(4,5)P 2 and PI (3, 4)P 2 (14), and a peptide with a charge equal to that of MARCKS (151-175), Lys-13, is also found to bind to membranes containing PI(4,5)P 2 with a similar  1) (lower traces). The total lipid concentration was ϳ40 mM, and the peptides were added externally to concentrations between 30 and 100 M. Spectra shown are for peptides labeled with a single R1 side chain (A) and peptides double-labeled with the R1 side chain (B). Given their high membrane affinity, these spin-labeled peptides will be entirely membrane-associated at the lipid concentrations used here (14). An aqueous peptide population would also be readily apparent from the EPR spectra of these spin-labeled peptides, and none is detected (38).

TABLE II
Power saturation and depths of spin-labeled MARCKS  bound to PI(4,5)P The depth parameter, ⌽, was obtained using 20 mM NiAA as the relaxation reagent, and the distances were estimated using a calibration published previously (28). The depth, d, is the position of the nitroxide relative to the lipid phosphate, where negative numbers indicate a location on the aqueous side of the phosphate, and positive numbers indicate a location on the hydrocarbon side. Ϫ1.4 Ϫ2 Ϫ4 S12R1 1.0 ϩ10 ϩ7 S19R1 0.5 ϩ7 ϩ6 a These depths were reported previously and were obtained in PC:PS using 40 mM NiEDDA as the relaxation reagent (22). The value given for S19R in PC:PS is an average depth obtained for the nearby residues 21 and 22. The error in the depth measurement for these peptides is approximately Ϯ2 Å. affinity. 2 Thus, MARCKS is likely to sequester other phosphoinositides based largely on their valence. It should be noted that interactions with proxyl-PIP 2 are not seen for all basic peptides. When we titrated PC:PIP 2 -proxyl membranes with pentalysine under the same conditions used here for MARCKS (151-175), there was no significant change in line shape and no evidence for a high affinity interaction (data not shown), consistent with the finding that pentalysine does not bind to PC: PI(4,5)P 2 (39). Thus, although the N-terminal end of the MARCKS effector domain begins with a Lys 5 sequence, this sequence alone is not sufficient to sequester PI(4,5)P 2 . Again, this is consistent with the idea that the ability of MARCKS effector domain to sequester PI(4,5)P 2 is driven by electrostatic interactions. Electrostatic fields are additive, and the distribution of ions around a charged site will depend upon the exponent of the valence of the site. As a result, peptides with a larger net positive charge will have a greater ability to alter the lateral distribution of negatively charged lipids such as PI(4,5)P 2 . Discrete binding sites for PI(4,5)P 2 do not exist on MARCKS (151-175); rather, it is the sum of positive charge and proximity to the membrane interface that are responsible for the observed MARCKS-PI(4,5)P 2 interaction. The strength of the electrostatic interaction of Lys 5 alone is insufficient to sequester this lipid.
In summary, a spin-labeled derivative of PI(4,5)P 2 has been synthesized and shown to report interactions between PI(4,5)P 2 and molecules that are known to bind the PI(4,5)P 2 head group within the plane of the bilayer. This probe is sensitive to changes in label motion which result from interactions at the membrane interface, and it can be used to determine the stoichiometry of protein-PI(4,5)P 2 interactions. The probe is also sensitive to local clustering, such as might be found in a PI(4,5)P 2 -rich domain. Data obtained using this probe indicate that a peptide derived from the effector domain of MARCKS interacts with PI(4,5)P 2 with a stoichiometry that is greater than 1:1. These data support the hypothesis that MARCKS functions to sequester PI(4,5)P 2 within the plane of the bilayer.